Spread Duration: Sensitivity Without a Default Forecast

Name the spread being changed

Spread duration begins with a pricing relationship, not just a field called spread. A model might add a constant spread to a spot curve, compare a bond yield with a benchmark yield, or incorporate an option-adjusted spread in a more elaborate valuation. Each construction has different held-fixed assumptions. The sensitivity should carry that definition wherever it appears, including portfolio summaries and scenario tables that omit the underlying instrument detail.

CFA Institute describes spread duration as price sensitivity to a change in yield spread. BIS research on benchmark choice shows why the comparison rate also matters. A measured gap can change when the benchmark changes, so a sensitivity calculation should distinguish an economic spread shock from a redefinition of the reference. Two reports using different benchmarks need not reconcile simply because both express their inputs in basis points.

Keep proportional and currency sensitivity separate

Using the common positive-duration convention, the approximation is change in value divided by base value approximately equal to -D_spread * change in spread. The spread change is an annual decimal rate. A one-basis-point movement is 0.0001. Multiplying base value by D_spread and 0.0001 gives the approximate currency loss magnitude for a one-basis-point widening, rather than a signed gain for every possible movement.

The base must be the value used in the sensitivity definition, not an unrelated face amount. For a hypothetical USD 250,000 base value and spread duration of 4.0 years, the one-basis-point widening estimate is USD 100 of loss magnitude. The independent scaling is 250,000 * 4 * 0.0001. Doubling the position doubles currency sensitivity while leaving the proportional duration unchanged under the same pricing assumptions.

Work through a widening with explicit signs

Apply an assumed 40-basis-point widening to that hypothetical position, keeping the benchmark curve and promised payments unchanged. The decimal spread move is +0.0040. The estimated proportional price change is -4.0 * 0.0040 = -0.0160, or -1.60%. Multiplying by USD 250,000 gives -USD 4,000. The same answer follows from 40 basis points times the USD 100 loss magnitude per basis point.

This is a first-order estimate using assumed sensitivity, not a calibrated instrument value or an executable price. It excludes curvature, trading costs and default. Under the same local approximation, a narrowing reverses the sign, but exact up and down repricing need not have equal magnitudes. Retaining the unrounded calculation before displaying dollars helps distinguish rounding differences from a genuine disagreement about definitions or model behavior.

Combine rate and spread effects only under one framework

A two-factor approximation can separate benchmark-rate and spread contributions if both sensitivities come from a compatible valuation. Add an assumed rate duration of 5.0 years and a benchmark decline of 20 basis points to the example. Its rate contribution is -250,000 * 5 * (-0.0020) = +USD 2,500. Combining that with the spread estimate gives approximately -USD 1,500, or -0.60% of base value.

The offset does not mean that spread risk disappeared. It means the selected rate movement compensated for part of the selected spread effect in this particular calculation. A review should show each contribution before the net result. Exact repricing may also contain cross-effects and curvature that are absent from the linear sum. If the sensitivities were measured under incompatible conventions, the apparent decomposition would not be interpretable.

Do not confuse an absolute shock with a relative one

A fixed number of basis points is an absolute spread change. A fixed percentage of the starting spread is a relative change. The two answer different scenario questions. For example, increasing an assumed spread from 100 to 150 basis points is a 50-basis-point move and a 50% relative increase. Moving from 400 to 450 basis points is the same absolute move but a 12.5% relative increase.

An absolute-shock comparison isolates sensitivity to the same rate increment; a relative-shock comparison scales the increment with the starting level. Neither is automatically the correct stress. State the purpose before choosing one, and avoid interpreting a result generated by one convention as evidence about the other. A sensitivity number by itself contains no probability model telling a reader which spread movement is likely or how severe conditions might become.

Identify when cash-flow uncertainty takes over

As default becomes central, changing only discount rates while retaining every promised coupon can become a poor description of value. Recovery assumptions and the timing of impaired payments may matter more than a small spread perturbation. Callable bonds add another complication because exercise behavior can change as rates or spreads move. A duration computed from fixed payments cannot automatically capture either mechanism merely by using a larger spread.

A numerical review should therefore separate a local derivative check from a broader economic suitability check. Agreement between two implementations of the same formula can establish that they compute the same approximation, yet leave the cash-flow assumptions inappropriate. Present both the approximation residual under exact model repricing and the features excluded by the model. Neither quantity substitutes for the other, and neither proves that a real issuer will pay.

Read single-yield tools conservatively

In a simplified fixed-payment model where total yield equals a fixed benchmark plus an additive spread, changing the spread changes the total yield one for one. In that restricted construction, spread sensitivity can coincide with yield sensitivity. That equivalence is an assumption of the model, not a general identity across bond analytics. Curve-based spreads, optionality and changing expected cash flows can break the relationship.

The parent tools calculate regular noncallable fixed-coupon bullet values using one annual nominal yield. They do not calibrate credit curves, estimate option-adjusted spreads or model default. A hypothetical yield shock may illustrate discounting, but its interpretation must remain conditional. Use the educational arithmetic to check units and attribution, not to convert a narrow present-value calculation into a comprehensive assessment of credit risk or a suggested position size.

Primary references

1. CFA Institute: Fixed-Income Active Management, Credit Strategies
2. BIS: Compressed Credit Spreads and the Quest for a Risk-Free Rate